AI agents have officially graduated. Once confined to the laptops of developers and experimental coding sandboxes, they are rapidly shifting into customer-facing products and mission-critical enterprise workflows. But as agents move into production, enterprises are discovering a new headache: crippling vendor lock-in and skyrocketing infrastructure bills.
Today enterprise AI infrastructure startup TrueFoundry is looking to capitalize on that friction. The company announced the launch of TrueForge, an open-source, vendor-neutral agent harness designed to go toe-to-toe with proprietary platforms like Claude Managed Agents.
The pitch is straightforward but ambitious: give enterprise teams the ability to build, deploy, debug, and govern production AI agents on any model or Model Context Protocol (MCP) server, while slashing operating costs by up to 50%.
The launch comes at a critical inflection point for the AI industry. Open-weight models, like the recently highly-touted GLM-5.2l are aggressively closing the performance gap with frontier proprietary models, and at a fraction of the price. Yet, the orchestration layer remains a walled garden. Most managed agent platforms currently force enterprises into a single vendor’s ecosystem, marrying them to specific models, infrastructure, and pricing tiers.
TrueFoundry is betting that enterprises will want to own that underlying orchestration layer rather than rent it from the LLM providers themselves.
“The biggest lesson from building AI infrastructure at Meta was that control and convenience aren’t opposites, with the right platform, you get both,” Nikunj Bajaj, co-founder and CEO of TrueFoundry, said in a statement. “Outside of big tech, teams are still forced to pick one the moment they hit production. TrueForge closes that gap: a world-class agent-building experience, open source, on any model, with enterprise governance built in at a fraction of the cost.”
Operating autonomous agents at scale requires more than just a clever prompt. Companies have to manage persistent sessions, securely handle tool credentials, spin up execution sandboxes, maintain context windows, enforce human-in-the-loop approvals, while keeping an eye on spend.
To handle this, TrueForge doesn’t just act as a standalone harness; it directly integrates with TrueFoundry’s existing AI Gateway, which the company claims already processes over a trillion tokens a day for its enterprise clients. By routing every model call and MCP interaction through this gateway, IT teams can enforce budgets, set rate limits, and apply guardrails centrally, rather than hoping individual developers are adhering to best practices.
Under the hood, the open-source platform is model-agnostic. It ships with out-of-the-box support for OpenAI, Anthropic, and over 20 other models. It also includes more than 40 built-in tools, web search powered by Tavily, generative UI capabilities, and sandboxed execution for secure code running. Teams can route specific tasks to whatever model fits the cost, latency, or quality requirements of that exact moment.
The platform is already seeing adoption among notable enterprise players. Companies like Automatiq and data infrastructure giant NetApp are utilizing TrueForge for their agentic workloads. TrueFoundry is essentially eating its own dog food; the company’s conversational interface, Ask TFY, runs entirely on the TrueForge harness.
“TrueFoundry has become the central platform in IT where agentic apps and agents are routed through,” said Robert Rubin, Senior Director of Platform Engineering at NetApp. “We are onboarding more users and teams every day, and it’s fundamentally changed how quickly we can get an agent from an idea to something running at scale.”
While TrueForge is available today on GitHub as a free, open-source release for teams bringing their own API keys and infrastructure, TrueFoundry isn’t ignoring the managed services market. Alongside the open-source drop, the company is rolling out a hosted, pay-per-usage version for teams that want the deployment convenience without playing system administrator.
As the enterprise AI narrative shifts from “what can these models do?” to “how do we control and afford them at scale?”, the infrastructure layer is becoming the ultimate strategic control point. By commoditizing the model connection and open-sourcing the agent harness, TrueFoundry is making a clear play to be the command center for the next era of enterprise AI.
Try today at https://github.com/truefoundry/trueforge.
